Filtered by vendor Linux
Subscribe
Total
19806 CVE
| CVE | Vendors | Products | Updated | CVSS v2 | CVSS v3 |
|---|---|---|---|---|---|
| CVE-2026-43453 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 7.1 HIGH |
| In the Linux kernel, the following vulnerability has been resolved: netfilter: nft_set_pipapo: fix stack out-of-bounds read in pipapo_drop() pipapo_drop() passes rulemap[i + 1].n to pipapo_unmap() as the to_offset argument on every iteration, including the last one where i == m->field_count - 1. This reads one element past the end of the stack-allocated rulemap array (declared as rulemap[NFT_PIPAPO_MAX_FIELDS] with NFT_PIPAPO_MAX_FIELDS == 16). Although pipapo_unmap() returns early when is_last is true without using the to_offset value, the argument is evaluated at the call site before the function body executes, making this a genuine out-of-bounds stack read confirmed by KASAN: BUG: KASAN: stack-out-of-bounds in pipapo_drop+0x50c/0x57c [nf_tables] Read of size 4 at addr ffff8000810e71a4 This frame has 1 object: [32, 160) 'rulemap' The buggy address is at offset 164 -- exactly 4 bytes past the end of the rulemap array. Pass 0 instead of rulemap[i + 1].n on the last iteration to avoid the out-of-bounds read. | |||||
| CVE-2026-43235 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: media: iris: Add missing platform data entries for SM8750 Two platform-data fields for SM8750 were missed: - get_vpu_buffer_size = iris_vpu33_buf_size Without this, the driver fails to allocate the required internal buffers, leading to basic decode/encode failures during session bring-up. - max_core_mbps = ((7680 * 4320) / 256) * 60 Without this capability exposed, capability checks are incomplete and v4l2-compliance for encoder fails. | |||||
| CVE-2026-43335 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: interconnect: qcom: sm8450: Fix NULL pointer dereference in icc_link_nodes() The change to dynamic IDs for SM8450 platform interconnects left two links unconverted, fix it to avoid the NULL pointer dereference in runtime, when a pointer to a destination interconnect is not valid: Unable to handle kernel NULL pointer dereference at virtual address 0000000000000008 <...> Call trace: icc_link_nodes+0x3c/0x100 (P) qcom_icc_rpmh_probe+0x1b4/0x528 platform_probe+0x64/0xc0 really_probe+0xc4/0x2a8 __driver_probe_device+0x80/0x140 driver_probe_device+0x48/0x170 __device_attach_driver+0xc0/0x148 bus_for_each_drv+0x88/0xf0 __device_attach+0xb0/0x1c0 device_initial_probe+0x58/0x68 bus_probe_device+0x40/0xb8 deferred_probe_work_func+0x90/0xd0 process_one_work+0x15c/0x3c0 worker_thread+0x2e8/0x400 kthread+0x150/0x208 ret_from_fork+0x10/0x20 Code: 900310f4 911d6294 91008280 94176078 (f94002a0) ---[ end trace 0000000000000000 ]--- Kernel panic - not syncing: Oops: Fatal exception | |||||
| CVE-2026-43330 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 7.8 HIGH |
| In the Linux kernel, the following vulnerability has been resolved: crypto: caam - fix overflow on long hmac keys When a key longer than block size is supplied, it is copied and then hashed into the real key. The memory allocated for the copy needs to be rounded to DMA cache alignment, as otherwise the hashed key may corrupt neighbouring memory. The copying is performed using kmemdup, however this leads to an overflow: reading more bytes (aligned_len - keylen) from the keylen source buffer. Fix this by replacing kmemdup with kmalloc, followed by memcpy. | |||||
| CVE-2026-43306 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: bpf: crypto: Use the correct destructor kfunc type With CONFIG_CFI enabled, the kernel strictly enforces that indirect function calls use a function pointer type that matches the target function. I ran into the following type mismatch when running BPF self-tests: CFI failure at bpf_obj_free_fields+0x190/0x238 (target: bpf_crypto_ctx_release+0x0/0x94; expected type: 0xa488ebfc) Internal error: Oops - CFI: 00000000f2008228 [#1] SMP ... As bpf_crypto_ctx_release() is also used in BPF programs and using a void pointer as the argument would make the verifier unhappy, add a simple stub function with the correct type and register it as the destructor kfunc instead. | |||||
| CVE-2026-43450 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 7.1 HIGH |
| In the Linux kernel, the following vulnerability has been resolved: netfilter: nfnetlink_cthelper: fix OOB read in nfnl_cthelper_dump_table() nfnl_cthelper_dump_table() has a 'goto restart' that jumps to a label inside the for loop body. When the "last" helper saved in cb->args[1] is deleted between dump rounds, every entry fails the (cur != last) check, so cb->args[1] is never cleared. The for loop finishes with cb->args[0] == nf_ct_helper_hsize, and the 'goto restart' jumps back into the loop body bypassing the bounds check, causing an 8-byte out-of-bounds read on nf_ct_helper_hash[nf_ct_helper_hsize]. The 'goto restart' block was meant to re-traverse the current bucket when "last" is no longer found, but it was placed after the for loop instead of inside it. Move the block into the for loop body so that the restart only occurs while cb->args[0] is still within bounds. BUG: KASAN: slab-out-of-bounds in nfnl_cthelper_dump_table+0x9f/0x1b0 Read of size 8 at addr ffff888104ca3000 by task poc_cthelper/131 Call Trace: nfnl_cthelper_dump_table+0x9f/0x1b0 netlink_dump+0x333/0x880 netlink_recvmsg+0x3e2/0x4b0 sock_recvmsg+0xde/0xf0 __sys_recvfrom+0x150/0x200 __x64_sys_recvfrom+0x76/0x90 do_syscall_64+0xc3/0x6e0 Allocated by task 1: __kvmalloc_node_noprof+0x21b/0x700 nf_ct_alloc_hashtable+0x65/0xd0 nf_conntrack_helper_init+0x21/0x60 nf_conntrack_init_start+0x18d/0x300 nf_conntrack_standalone_init+0x12/0xc0 | |||||
| CVE-2026-43415 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 4.7 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: scsi: ufs: core: Fix SError in ufshcd_rtc_work() during UFS suspend In __ufshcd_wl_suspend(), cancel_delayed_work_sync() is called to cancel the UFS RTC work, but it is placed after ufshcd_vops_suspend(hba, pm_op, POST_CHANGE). This creates a race condition where ufshcd_rtc_work() can still be running while ufshcd_vops_suspend() is executing. When UFSHCD_CAP_CLK_GATING is not supported, the condition !hba->clk_gating.active_reqs is always true, causing ufshcd_update_rtc() to be executed. Since ufshcd_vops_suspend() typically performs clock gating operations, executing ufshcd_update_rtc() at that moment triggers an SError. The kernel panic trace is as follows: Kernel panic - not syncing: Asynchronous SError Interrupt Call trace: dump_backtrace+0xec/0x128 show_stack+0x18/0x28 dump_stack_lvl+0x40/0xa0 dump_stack+0x18/0x24 panic+0x148/0x374 nmi_panic+0x3c/0x8c arm64_serror_panic+0x64/0x8c do_serror+0xc4/0xc8 el1h_64_error_handler+0x34/0x4c el1h_64_error+0x68/0x6c el1_interrupt+0x20/0x58 el1h_64_irq_handler+0x18/0x24 el1h_64_irq+0x68/0x6c ktime_get+0xc4/0x12c ufshcd_mcq_sq_stop+0x4c/0xec ufshcd_mcq_sq_cleanup+0x64/0x1dc ufshcd_clear_cmd+0x38/0x134 ufshcd_issue_dev_cmd+0x298/0x4d0 ufshcd_exec_dev_cmd+0x1a4/0x1c4 ufshcd_query_attr+0xbc/0x19c ufshcd_rtc_work+0x10c/0x1c8 process_scheduled_works+0x1c4/0x45c worker_thread+0x32c/0x3e8 kthread+0x120/0x1d8 ret_from_fork+0x10/0x20 Fix this by moving cancel_delayed_work_sync() before the call to ufshcd_vops_suspend(hba, pm_op, PRE_CHANGE), ensuring the UFS RTC work is fully completed or cancelled at that point. | |||||
| CVE-2026-43401 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: cpufreq: intel_pstate: Fix NULL pointer dereference in update_cpu_qos_request() The update_cpu_qos_request() function attempts to initialize the 'freq' variable by dereferencing 'cpudata' before verifying if the 'policy' is valid. This issue occurs on systems booted with the "nosmt" parameter, where all_cpu_data[cpu] is NULL for the SMT sibling threads. As a result, any call to update_qos_requests() will result in a NULL pointer dereference as the code will attempt to access pstate.turbo_freq using the NULL cpudata pointer. Also, pstate.turbo_freq may be updated by intel_pstate_get_hwp_cap() after initializing the 'freq' variable, so it is better to defer the 'freq' until intel_pstate_get_hwp_cap() has been called. Fix this by deferring the 'freq' assignment until after the policy and driver_data have been validated. [ rjw: Added one paragraph to the changelog ] | |||||
| CVE-2026-43290 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 7.8 HIGH |
| In the Linux kernel, the following vulnerability has been resolved: media: uvcvideo: Return queued buffers on start_streaming() failure Return buffers if streaming fails to start due to uvc_pm_get() error. This bug may be responsible for a warning I got running while :; do yavta -c3 /dev/video0; done on an xHCI controller which failed under this workload. I had no luck reproducing this warning again to confirm. xhci_hcd 0000:09:00.0: HC died; cleaning up usb 13-2: USB disconnect, device number 2 WARNING: CPU: 2 PID: 29386 at drivers/media/common/videobuf2/videobuf2-core.c:1803 vb2_start_streaming+0xac/0x120 | |||||
| CVE-2026-43190 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 8.2 HIGH |
| In the Linux kernel, the following vulnerability has been resolved: netfilter: xt_tcpmss: check remaining length before reading optlen Quoting reporter: In net/netfilter/xt_tcpmss.c (lines 53-68), the TCP option parser reads op[i+1] directly without validating the remaining option length. If the last byte of the option field is not EOL/NOP (0/1), the code attempts to index op[i+1]. In the case where i + 1 == optlen, this causes an out-of-bounds read, accessing memory past the optlen boundary (either reading beyond the stack buffer _opt or the following payload). | |||||
| CVE-2026-43227 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: clocksource/drivers/sh_tmu: Always leave device running after probe The TMU device can be used as both a clocksource and a clockevent provider. The driver tries to be smart and power itself on and off, as well as enabling and disabling its clock when it's not in operation. This behavior is slightly altered if the TMU is used as an early platform device in which case the device is left powered on after probe, but the clock is still enabled and disabled at runtime. This has worked for a long time, but recent improvements in PREEMPT_RT and PROVE_LOCKING have highlighted an issue. As the TMU registers itself as a clockevent provider, clockevents_register_device(), it needs to use raw spinlocks internally as this is the context of which the clockevent framework interacts with the TMU driver. However in the context of holding a raw spinlock the TMU driver can't really manage its power state or clock with calls to pm_runtime_*() and clk_*() as these calls end up in other platform drivers using regular spinlocks to control power and clocks. This mix of spinlock contexts trips a lockdep warning. ============================= [ BUG: Invalid wait context ] 6.18.0-arm64-renesas-09926-gee959e7c5e34 #1 Not tainted ----------------------------- swapper/0/0 is trying to lock: ffff000008c9e180 (&dev->power.lock){-...}-{3:3}, at: __pm_runtime_resume+0x38/0x88 other info that might help us debug this: context-{5:5} 1 lock held by swapper/0/0: ccree e6601000.crypto: ARM CryptoCell 630P Driver: HW version 0xAF400001/0xDCC63000, Driver version 5.0 #0: ffff8000817ec298 ccree e6601000.crypto: ARM ccree device initialized (tick_broadcast_lock){-...}-{2:2}, at: __tick_broadcast_oneshot_control+0xa4/0x3a8 stack backtrace: CPU: 0 UID: 0 PID: 0 Comm: swapper/0 Not tainted 6.18.0-arm64-renesas-09926-gee959e7c5e34 #1 PREEMPT Hardware name: Renesas Salvator-X 2nd version board based on r8a77965 (DT) Call trace: show_stack+0x14/0x1c (C) dump_stack_lvl+0x6c/0x90 dump_stack+0x14/0x1c __lock_acquire+0x904/0x1584 lock_acquire+0x220/0x34c _raw_spin_lock_irqsave+0x58/0x80 __pm_runtime_resume+0x38/0x88 sh_tmu_clock_event_set_oneshot+0x84/0xd4 clockevents_switch_state+0xfc/0x13c tick_broadcast_set_event+0x30/0xa4 __tick_broadcast_oneshot_control+0x1e0/0x3a8 tick_broadcast_oneshot_control+0x30/0x40 cpuidle_enter_state+0x40c/0x680 cpuidle_enter+0x30/0x40 do_idle+0x1f4/0x280 cpu_startup_entry+0x34/0x40 kernel_init+0x0/0x130 do_one_initcall+0x0/0x230 __primary_switched+0x88/0x90 For non-PREEMPT_RT builds this is not really an issue, but for PREEMPT_RT builds where normal spinlocks can sleep this might be an issue. Be cautious and always leave the power and clock running after probe. | |||||
| CVE-2026-43417 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: sched/mmcid: Handle vfork()/CLONE_VM correctly Matthieu and Jiri reported stalls where a task endlessly loops in mm_get_cid() when scheduling in. It turned out that the logic which handles vfork()'ed tasks is broken. It is invoked when the number of tasks associated to a process is smaller than the number of MMCID users. It then walks the task list to find the vfork()'ed task, but accounts all the already processed tasks as well. If that double processing brings the number of to be handled tasks to 0, the walk stops and the vfork()'ed task's CID is not fixed up. As a consequence a subsequent schedule in fails to acquire a (transitional) CID and the machine stalls. Cure this by removing the accounting condition and make the fixup always walk the full task list if it could not find the exact number of users in the process' thread list. | |||||
| CVE-2026-43248 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 7.8 HIGH |
| In the Linux kernel, the following vulnerability has been resolved: vhost: move vdpa group bound check to vhost_vdpa Remove duplication by consolidating these here. This reduces the posibility of a parent driver missing them. While we're at it, fix a bug in vdpa_sim where a valid ASID can be assigned to a group equal to ngroups, causing an out of bound write. | |||||
| CVE-2026-43319 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: spi: spidev: fix lock inversion between spi_lock and buf_lock The spidev driver previously used two mutexes, spi_lock and buf_lock, but acquired them in different orders depending on the code path: write()/read(): buf_lock -> spi_lock ioctl(): spi_lock -> buf_lock This AB-BA locking pattern triggers lockdep warnings and can cause real deadlocks: WARNING: possible circular locking dependency detected spidev_ioctl() -> mutex_lock(&spidev->buf_lock) spidev_sync_write() -> mutex_lock(&spidev->spi_lock) *** DEADLOCK *** The issue is reproducible with a simple userspace program that performs write() and SPI_IOC_WR_MAX_SPEED_HZ ioctl() calls from separate threads on the same spidev file descriptor. Fix this by simplifying the locking model and removing the lock inversion entirely. spidev_sync() no longer performs any locking, and all callers serialize access using spi_lock. buf_lock is removed since its functionality is fully covered by spi_lock, eliminating the possibility of lock ordering issues. This removes the lock inversion and prevents deadlocks without changing userspace ABI or behaviour. | |||||
| CVE-2026-43196 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 7.8 HIGH |
| In the Linux kernel, the following vulnerability has been resolved: soc: ti: pruss: Fix double free in pruss_clk_mux_setup() In the pruss_clk_mux_setup(), the devm_add_action_or_reset() indirectly calls pruss_of_free_clk_provider(), which calls of_node_put(clk_mux_np) on the error path. However, after the devm_add_action_or_reset() returns, the of_node_put(clk_mux_np) is called again, causing a double free. Fix by returning directly, to avoid the duplicate of_node_put(). | |||||
| CVE-2026-43340 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: comedi: Reinit dev->spinlock between attachments to low-level drivers `struct comedi_device` is the main controlling structure for a COMEDI device created by the COMEDI subsystem. It contains a member `spinlock` containing a spin-lock that is initialized by the COMEDI subsystem, but is reserved for use by a low-level driver attached to the COMEDI device (at least since commit 25436dc9d84f ("Staging: comedi: remove RT code")). Some COMEDI devices (those created on initialization of the COMEDI subsystem when the "comedi.comedi_num_legacy_minors" parameter is non-zero) can be attached to different low-level drivers over their lifetime using the `COMEDI_DEVCONFIG` ioctl command. This can result in inconsistent lock states being reported when there is a mismatch in the spin-lock locking levels used by each low-level driver to which the COMEDI device has been attached. Fix it by reinitializing `dev->spinlock` before calling the low-level driver's `attach` function pointer if `CONFIG_LOCKDEP` is enabled. | |||||
| CVE-2026-43404 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: mm: Fix a hmm_range_fault() livelock / starvation problem If hmm_range_fault() fails a folio_trylock() in do_swap_page, trying to acquire the lock of a device-private folio for migration, to ram, the function will spin until it succeeds grabbing the lock. However, if the process holding the lock is depending on a work item to be completed, which is scheduled on the same CPU as the spinning hmm_range_fault(), that work item might be starved and we end up in a livelock / starvation situation which is never resolved. This can happen, for example if the process holding the device-private folio lock is stuck in migrate_device_unmap()->lru_add_drain_all() sinc lru_add_drain_all() requires a short work-item to be run on all online cpus to complete. A prerequisite for this to happen is: a) Both zone device and system memory folios are considered in migrate_device_unmap(), so that there is a reason to call lru_add_drain_all() for a system memory folio while a folio lock is held on a zone device folio. b) The zone device folio has an initial mapcount > 1 which causes at least one migration PTE entry insertion to be deferred to try_to_migrate(), which can happen after the call to lru_add_drain_all(). c) No or voluntary only preemption. This all seems pretty unlikely to happen, but indeed is hit by the "xe_exec_system_allocator" igt test. Resolve this by waiting for the folio to be unlocked if the folio_trylock() fails in do_swap_page(). Rename migration_entry_wait_on_locked() to softleaf_entry_wait_unlock() and update its documentation to indicate the new use-case. Future code improvements might consider moving the lru_add_drain_all() call in migrate_device_unmap() to be called *after* all pages have migration entries inserted. That would eliminate also b) above. v2: - Instead of a cond_resched() in hmm_range_fault(), eliminate the problem by waiting for the folio to be unlocked in do_swap_page() (Alistair Popple, Andrew Morton) v3: - Add a stub migration_entry_wait_on_locked() for the !CONFIG_MIGRATION case. (Kernel Test Robot) v4: - Rename migrate_entry_wait_on_locked() to softleaf_entry_wait_on_locked() and update docs (Alistair Popple) v5: - Add a WARN_ON_ONCE() for the !CONFIG_MIGRATION version of softleaf_entry_wait_on_locked(). - Modify wording around function names in the commit message (Andrew Morton) (cherry picked from commit a69d1ab971a624c6f112cea61536569d579c3215) | |||||
| CVE-2026-43459 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 7.3 HIGH |
| In the Linux kernel, the following vulnerability has been resolved: ASoC: soc-core: flush delayed work before removing DAIs and widgets When a sound card is unbound while a PCM stream is open, a use-after-free can occur in snd_soc_dapm_stream_event(), called from the close_delayed_work workqueue handler. During unbind, snd_soc_unbind_card() flushes delayed work and then calls soc_cleanup_card_resources(). Inside cleanup, snd_card_disconnect_sync() releases all PCM file descriptors, and the resulting PCM close path can call snd_soc_dapm_stream_stop() which schedules new delayed work with a pmdown_time timer delay. Since this happens after the flush in snd_soc_unbind_card(), the new work is not caught. soc_remove_link_components() then frees DAPM widgets before this work fires, leading to the use-after-free. The existing flush in soc_free_pcm_runtime() also cannot help as it runs after soc_remove_link_components() has already freed the widgets. Add a flush in soc_cleanup_card_resources() after snd_card_disconnect_sync() (after which no new PCM closes can schedule further delayed work) and before soc_remove_link_dais() and soc_remove_link_components() (which tear down the structures the delayed work accesses). | |||||
| CVE-2026-43347 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 7.5 HIGH |
| In the Linux kernel, the following vulnerability has been resolved: arm64: dts: qcom: monaco: Reserve full Gunyah metadata region We observe spurious "Synchronous External Abort" exceptions (ESR=0x96000010) and kernel crashes on Monaco-based platforms. These faults are caused by the kernel inadvertently accessing hypervisor-owned memory that is not properly marked as reserved. >From boot log, The Qualcomm hypervisor reports the memory range at 0x91a80000 of size 0x80000 (512 KiB) as hypervisor-owned: qhee_hyp_assign_remove_memory: 0x91a80000/0x80000 -> ret 0 However, the EFI memory map provided by firmware only reserves the subrange 0x91a40000–0x91a87fff (288 KiB). The remaining portion (0x91a88000–0x91afffff) is incorrectly reported as conventional memory (from efi debug): efi: 0x000091a40000-0x000091a87fff [Reserved...] efi: 0x000091a88000-0x0000938fffff [Conventional...] As a result, the allocator may hand out PFNs inside the hypervisor owned region, causing fatal aborts when the kernel accesses those addresses. Add a reserved-memory carveout for the Gunyah hypervisor metadata at 0x91a80000 (512 KiB) and mark it as no-map so Linux does not map or allocate from this area. For the record: Hyp version: gunyah-e78adb36e debug (2025-11-17 05:38:05 UTC) UEFI Ver: 6.0.260122.BOOT.MXF.1.0.c1-00449-KODIAKLA-1 | |||||
| CVE-2026-43161 | 1 Linux | 1 Linux Kernel | 2026-11-11 | N/A | 5.5 MEDIUM |
| In the Linux kernel, the following vulnerability has been resolved: iommu/vt-d: Skip dev-iotlb flush for inaccessible PCIe device without scalable mode PCIe endpoints with ATS enabled and passed through to userspace (e.g., QEMU, DPDK) can hard-lock the host when their link drops, either by surprise removal or by a link fault. Commit 4fc82cd907ac ("iommu/vt-d: Don't issue ATS Invalidation request when device is disconnected") adds pci_dev_is_disconnected() to devtlb_invalidation_with_pasid() so ATS invalidation is skipped only when the device is being safely removed, but it applies only when Intel IOMMU scalable mode is enabled. With scalable mode disabled or unsupported, a system hard-lock occurs when a PCIe endpoint's link drops because the Intel IOMMU waits indefinitely for an ATS invalidation that cannot complete. Call Trace: qi_submit_sync qi_flush_dev_iotlb __context_flush_dev_iotlb.part.0 domain_context_clear_one_cb pci_for_each_dma_alias device_block_translation blocking_domain_attach_dev iommu_deinit_device __iommu_group_remove_device iommu_release_device iommu_bus_notifier blocking_notifier_call_chain bus_notify device_del pci_remove_bus_device pci_stop_and_remove_bus_device pciehp_unconfigure_device pciehp_disable_slot pciehp_handle_presence_or_link_change pciehp_ist Commit 81e921fd3216 ("iommu/vt-d: Fix NULL domain on device release") adds intel_pasid_teardown_sm_context() to intel_iommu_release_device(), which calls qi_flush_dev_iotlb() and can also hard-lock the system when a PCIe endpoint's link drops. Call Trace: qi_submit_sync qi_flush_dev_iotlb __context_flush_dev_iotlb.part.0 intel_context_flush_no_pasid device_pasid_table_teardown pci_pasid_table_teardown pci_for_each_dma_alias intel_pasid_teardown_sm_context intel_iommu_release_device iommu_deinit_device __iommu_group_remove_device iommu_release_device iommu_bus_notifier blocking_notifier_call_chain bus_notify device_del pci_remove_bus_device pci_stop_and_remove_bus_device pciehp_unconfigure_device pciehp_disable_slot pciehp_handle_presence_or_link_change pciehp_ist Sometimes the endpoint loses connection without a link-down event (e.g., due to a link fault); killing the process (virsh destroy) then hard-locks the host. Call Trace: qi_submit_sync qi_flush_dev_iotlb __context_flush_dev_iotlb.part.0 domain_context_clear_one_cb pci_for_each_dma_alias device_block_translation blocking_domain_attach_dev __iommu_attach_device __iommu_device_set_domain __iommu_group_set_domain_internal iommu_detach_group vfio_iommu_type1_detach_group vfio_group_detach_container vfio_group_fops_release __fput pci_dev_is_disconnected() only covers safe-removal paths; pci_device_is_present() tests accessibility by reading vendor/device IDs and internally calls pci_dev_is_disconnected(). On a ConnectX-5 (8 GT/s, x2) this costs ~70 µs. Since __context_flush_dev_iotlb() is only called on {attach,release}_dev paths (not hot), add pci_device_is_present() there to skip inaccessible devices and avoid the hard-lock. | |||||